On Thu, Apr 19, 2018 at 08:51:11AM +0200, Michal Hocko wrote: > On Thu 19-04-18 10:36:39, Wu Fengguang wrote: > > Hi Alexander, > > > > FYI this happens in mainline kernel 4.17.0-rc1. > > It dates back to at least v4.15. > > > > It occurs in 4 out of 4 boots. Here KVM has 1G memory. > > > > This high order allocation caused lots of noises in our boot testing. > > We could disable this device in our tests, but it would be great if > > there are better ways out. > > > > [ 75.039408] Product name: fake-design-for-testing > > [ 75.040995] fmc fake-design-for-testing-f001: Driver has no ID: matches > > all > > [ 75.042509] fmc_trivial: probe of fake-design-for-testing-f001 failed > > with error -95 > > [ 75.044323] fmc fake-design-for-testing-f001: Driver has no ID: matches > > all > > [ 75.045644] fmc_chardev fake-design-for-testing-f001: Created misc > > device "fake-design-for-testing-f001" > > [ 75.061570] swapper/0: page allocation failure: order:9, > > mode:0x14040c0(GFP_KERNEL|__GFP_COMP), nodemask=(null) > > Is there any reason why > > [ 75.063338] stm_register_device+0xf3/0x5c0: > > stm_register_device at > > drivers/hwtracing/stm/core.c:695 > > cannot use kvzalloc?
You're right, of course, I'll fix that. I can't see, though, how it ends up asking order==9 from the buddy allocator. The kmalloc() in question should be asking for ~512kB, theres should be smaller slabs for that. Regards, -- Alex
